The impact that antiretroviral therapy (ART) may have on the HIV epidemic in sub-Saharan Africa is highly reliant on the rate at which HIV-infected people initiate ART, HIV transmission probability while on ART [1,2], mortality rates while on ART, and, importantly, whether ART influences risky sexual behavior. Studies conducted in the US indicate that risky behavior may rise once ART becomes available . Similar observations have been reported in African countries , though some do not think that ART has impacted sexual behavior in Africa . Some even suggest that risky behavior may decrease among those on ART due to more intensive counseling . Previous studies in sub-Saharan Africa associating ART and sexual behavior have focused on HIV-infected people, and have reported results within a narrow time frame, for example, within 2 years of ART initiation.
Possible changes in sexual risk taking among HIV-uninfected people may, however, also impact HIV epidemiologic trends. To date, few long-term cohorts in sub-Saharan Africa have provided empirical evidence for sexual behavior trends following uptake of ART; none have provided evidence that expands more than 2 years after ART initiation, or after ART availability.
We examined self-reported evidence for changing sexual behavior after the introduction of ART in a cohort in rural Uganda in 2004. We compared sexual behavior before and after ART was available. We examined the behavior of both HIV-infected and uninfected people as the knowledge that ART is available could impact the behavior of both groups. Among those receiving ART, we compared behavior before and after ART initiation. In order to help assess the impact that ART may have on the HIV epidemic, we also briefly examined the proportion of HIV-infected people who have initiated ART, and the rate of acquiring new sexual partners.
In 1990, the Medical Research Council established a clinical cohort in rural SW Uganda [7,8]. Participants are adults aged 15 or more, identified from a large general population cohort with annual HIV serosurveys [9–11]. The cohort consists of a selection of HIV-prevalent cases identified in 1990, HIV-incident cases identified during subsequent survey rounds and age and sex-stratified randomly selected negative controls. Since 2004, the time when ART was introduced, 14-year-olds eligible for ART are also enrolled. We restricted analysis to those aged 15 and above in order to be consistent across time. A study physician sees participants every 3 months, takes a medical and sexual history, and performs a clinical examination. Among HIV-infected participants, CD4 cell count measurements are taken at the same time as the 3-monthly interviews.
Using interview data obtained at the 3-monthly visits, we describe trends in mean number of sex acts per week, mean number of current partners, mean number of casual partners, mean number of new partners in the past 3 months, proportion who used a condom at last sex with a casual partner, and proportion having had sex for money in the past 3 months. We conducted regression analyses among those receiving ART, using the above outcome indicators, whereas predictors were time before and after ART initiation, time squared in order to allow a nonlinear trend, and CD4 cell count. We considered CD4 cell count as a proxy for comparative well being. Confidence intervals were determined by Huber corrected robust standard errors , sometimes called sandwich estimators [13,14], in order to account for multiple measurements per individual.
Results were pooled across all people, as well as disaggregated by sex. Sexual partner turnover rates were examined separately, and stratified by age and sex. Since variability in sexual partner turnover rates, with some people having very high rates, is required to fit mathematical models to the HIV epidemic, we examined how variable these rates were in the cohort. The annual partner turnover rate was estimated by summing the new partners in the past 3 months across all four follow-ups in a year. In cases in which a participant had fewer than four follow-ups, we summed the new partners in the past 3 months across all follow-ups in the year and multiplied this number by [4/(number of follow-ups in the respective year)]. We describe annual sexual partner turnover using one record per year per sexually active participant.
The Science and Ethics Committee of the Uganda Virus Research Institute and the Uganda National Council for Science and Technology gave ethical approval for the study.
From January 2002 to March 2009, 669 people aged 15 and above ever participated in the clinical cohort. Of these, 55% were women, 16% died, 68% were HIV-infected, and 38% ever initiated ART. The age range at enrolment into the cohort was 15–85 years, with 26% under age 25, 54% between 25 and 44, and 20% aged 45 and above. Among the 455 HIV-infected people, 20% died and 55% ever initiated ART.
Antiretroviral therapy experience
The number of adults known to be HIV-infected in the general population cohort (GPC) from which the clinical cohort recruits participants ranged from 591 in 2004 to 684 in 2008. Among all known HIV-infected adults in the GPC, 7.5% were receiving ART in 2004. This rose to 15.6% in 2005, 19.5% in 2006, 23.3% in 2007, and 27.5% in 2008. Overall, 122 (95.3%) of the 128 in the clinical cohort who were eligible for ART had initiated it. Those who had not yet started treatment either refused to know their HIV status or refused to initiate ART. The HIV status is not known for about one-quarter of the people in the GPC. Although some of those with unknown status are likely to be HIV-infected as HIV prevalence in the GPC is 7–8%, none of those with unknown status had joined the clinical cohort or begun ART.
Annual sexual partner turnover
The number of annual interview records on sexual partners ranged from 240 in 2002 to a maximum of 352 in 2008, and 322 in 2009. Among sexually active people and across all years (2002–2009), no new partner in the past year was reported in 88.0% of the reports. There were some reports of many partners. For example, 4.3% of the reports indicated five or more new partners in the past year (2.2% of the reports were 5–9 new partners), 1.7% of the reports indicated 10 or more new partners in the past year, and 0.4% of the annual reports indicated more than 20 partners in the past year. The largest number of reported new partners in a year was 80.
We next assessed the maximum number of new partners reported by each cohort participant; that is, among participants with multiple years of reporting, we examined the maximum report in any given year. Between 2002 and 2009, 67.8% of the sexually active people never reported having had even one new partner, although 12.6% reported five or more new partners in at least 1 year, 5.0% reported 10 or more new partners in at least 1 year, and 1.4% reported more than 20 partners in at least 1 year.
A summary of partner turnover by age and sex is provided in Table 1. Sample size prevented further data disaggregation (Table 1). The mean partner turnover rate decreased with age. 18.7% of women under age 45 reported to have had 1–9 new partners in at least one of the years between 2002 and 2009 inclusive (see second to rightmost column of Table 1), and the mean number of new partners among those with 1–9 was 3.86. Among men under age 45, 39.3% reported to have had 1–9 new partners in at least one of the years between 2002 and 2009 inclusive, with a mean in that group of 3.97.
Antiretroviral therapy use and sexual behavior among HIV-infected
Sexual behavior 2 or more years after ART initiation was not significantly more risky than it was 2 or more years before ART for any indicator. Among continuous indicators, risky behavior dropped in the year before and around the timing of ART initiation, but by 2 or more years after ART it had risen again, though the changing trend was not always statistically significant (Table 2c). By contrast, among dichotomous indicators (used a condom at last sex with a casual partner, and had sex in the past 3 months for money or gifts), risky behavior declined throughout the period of analysis.
Among continuous variables, the pattern of sexual behavior with respect to the timing of ART initiation is consistent with people engaging in more risky behavior when they feel healthier. We assessed the relationship between sexual behavior and CD4 cell count, as a proxy for healthiness (Table 2d). There was a trend toward less risky behavior among some indicators as CD4 declined (number of sex acts in the past week and number of current partners), but not among others. For example, the highest percentage of those who reported to have had sex for money or gifts, or to have given money or gifts for sex, was found among those with a CD4 cell count below 50.
Regressions predicting sexual behavior by time with respect to ART initiation and by CD4 cell count were used to assess the impact of ART on sexual behavior, to some extent controlling for physical well being (Table 3a). The difference between these regressions and those performed to obtain P values in Table 2c is that in this set of regressions, CD4 was added to the model. CD4 cell count was not a statistically significant predictor of any outcome. However, in all outcomes except number of sex acts in the past week, the changing trend with respect to timing of ART initiation was statistically significant. For all models, the coefficient of time with respect to ART initiation was negative, whereas the coefficient of time squared was positive. This indicates that risky behavior declined initially and then either began to level off, or rose. In most cases, risky behavior rose after ART initiation (Fig. 1), though it did not rise to levels significantly higher than it was 2 or more years before ART initiation.
We next considered the possibility that ART may impact the sexual behavior of the partner of a person on ART. Between January 2004, when ART became available, and March 2009, there were 24 sexual partnerships in which both partners were enrolled in the cohort but only one was already on ART. By year between 2004 and 2008, the number of ART-discordant partnerships numbered 10, 10, 15, 11, and 9. About half of the ART-discordant partnerships became concordant with time. We assessed the coital frequency reported by the ART-naive partner in these partnerships, by time since ART initiation of the ART-experienced partner. There is no evidence of a change in reported sex acts per week by the partner of somebody on ART, though sample sizes were small. The mean number of reported sex acts per week reported by the ART-naive partner in the 12 months leading up to ART initiation of the ART-experienced partner was 1.32. In the 12 months after ART initiation, the naive partner reported 1.07 sex acts per week on average, but the difference between this and the pre-ART reports was not statistically significant (P = 0.577). The mean reported sex acts by the partner of the ART initiator at 1–2 years after ART initiation was 1.17, and it was 1.12 and 0.77 at 2–3 years and 3 or more years after ART initiation, respectively. Trends were not statistically significant.
Antiretroviral therapy availability and sexual behavior among HIV-uninfected
Among HIV-uninfected participants, we examined reported sexual behavior before and after ART became available. Among some indicators, there was no trend from 2002 until 2009. Among other indicators, risky behavior appeared to be declining before ART roll-out began in the area in 2004, and increasing after.
There was no trend in the reported number of current partners from 2002, through ART roll-out which began in 2004, and continuing through 2009 (Fig. 2). However, the mean number of casual partners fell from 0.02 in early 2002 to 0.01 by 2004 and then rose to 0.03 by late 2008; the change in trend was statistically significant (P = 0.030; Table 3b). Among both men and women, the number of sex acts per week rose after 2002 (Fig. 2). The regression results show that this indicator increased throughout the period of analysis (2002–2009), but rose more steeply in the later than in the earlier years (Table 3b). The mean reported number of new partners in the past 3 months was highly variable, particularly among men. However, it fell from 0.128 in the first quarter of 2002 to 0.014 and 0.021 in the second and third quarters of 2004. By the third and fourth quarters of 2008, four and a half years after ART roll-out began, the reported number of new partners in the past 3 months had risen to 0.087 (third quarter) and 0.200 (fourth quarter). Because of the high quarter to quarter variability in this indicator, the changing trend was not statistically significant (P = 0.058, Table 3).
Antiretroviral therapy initiation
In the general population from which the clinical cohort that we have described comes, there is intense follow-up in order to obtain HIV status. Nonetheless, in any given year, the cohort is comprised of about 25% with an unknown HIV status. None of those with unknown status begin ART. Outside of an intensely followed cohort, it is unlikely to expect less than the 25% that was found in this cohort to have an unknown HIV status. This has important implications when assessing the possible impact that ART may have on the course of the HIV epidemic as it constrains the plausible rate at which HIV-infected people can initiate ART.
Sexual partner turnover
Although 95.7% of responses indicated 4 or fewer new partners during the year, 12.6% of participants ever reported (in any of their annually summed responses) to have had five or more new partners during the year. Further, 1.4% of participants ever reported to have had 20 or more new partners during the year. Reproducing the epidemic through modeling generally requires an elevation in the mean number of new partners per year in the highest sexual activity group (those with more than 20 new partners in a year) over that reported. However, others have noted that community cohort surveys may miss highly sexually active people because of higher levels of migration and mobility among these groups . Highly active women are particularly likely to be under-represented because of biased reporting . We found, for example, that men reported a higher number of sex acts and a higher partner turnover than women (see Fig. 2). This discrepancy between reported partner turnover rates is a frequent phenomenon [16–18]. Yet, unless men are partnering outside of the cohort or exaggerate partner numbers, reported sexual activity should be similar in men and women.
Sexual behavior and antiretroviral therapy
During the first year after ART initiation, average risky behavior declined on every sexual behavior indicator that we examined, compared to behavior during the year preceding ART initiation (Table 2c). Although this difference in risky behavior was only statistically significant with regards to number of current partners (P = 0.002) and having used a condom at last sex with a casual partner (P < 0.001), the descriptive finding is consistent with those of Bunnell and colleagues , who found a decline in risky behavior 6 months after ART initiation. This decline may be due to a combination of feeling more sick shortly after ART initiation, and increased counseling intensity. However, the results of Bunnell et al. were truncated at 6 months after ART initiation. We found that by 2 or more years after ART initiation, risky behavior had again risen, though only to levels similar to that found before initiation.
The behavior of HIV-uninfected people may impact HIV epidemiologic trends as well. We cannot determine whether the reported behavior change among HIV-uninfected people occurred in response to ART availability or due to other temporal factors. Nevertheless, we have shown evidence that reported risky behavior as measured by some indicators was declining before 2004 when ART was introduced into the population, and increasing thereafter. This pattern held true particularly for reports on new partners in the past 3 months.
The findings of our data analysis are in line with literature sources. Among HIV-infected, some studies have found that ART use may be related to lower risk taking [6,19], which is supported by our data for the first year after ART initiation. As we have had a longer follow-up than other studies, however, we were able to document a rise in risky behavior, after the initial decline that occurred near the time that people began ART. Among HIV-uninfected, a study conducted outside of Africa showed a correlation between ART availability and higher risk taking behavior . We have also observed this, though not consistently among all indicators examined.
The most striking change in behavior trends is seen in the number of sex acts and the number of new partners among HIV-uninfected people after ART became available. As such, we intend to explore through mathematical modeling the impact of ART, alone and in combination with the potential rise in partner turnover rate, in another study.
Sexual behavior and CD4 cell count
We examined the association between sexual behavior and CD4 cell count in order to assess whether relative feeling of health may account for changes in sexual behavior around the timing of ART initiation. We found that the highest percentage of reports of exchanging money or gifts for sex was among those with CD4 cell counts below 50, and this was statistically significant (P < 0.01). This is a worrying observation that may indicate that poverty and the need to provide for food and medicines may drive people with advanced HIV disease to more risky sexual behavior at times when they are ill and highly infectious.
We chose to examine sexual behavior within this cohort because it is more intensely followed than the larger general population cohort from which participants in this smaller cohort are recruited. Our results depended on this because all results are based on time – either time since ART became available, or time since ART initiation. In the general population, behavior surveys are only conducted once per year, so time would not be as precise. However, there were limitations in choosing to present findings from this cohort. This is a clinical cohort and as such is not representative of the general population; 68% of the participants between 2002 and 2009 were HIV-infected. Most of our results are stratified by HIV infection; we present sexual behavior among HIV-infected and HIV-uninfected people separately. As such, an unrepresentatively high proportion of HIV-infected people in the cohort would not affect the generalizability of those results. The sexual partner turnover rates, however, are not stratified by HIV infection because there were too few data to stratify by age, sex, and HIV infection, after having already aggregated the 3-monthly reports into one report per year per person. We found that reported sexual partner turnover rates were higher by about 20% among HIV-infected than uninfected.
Following an initial decline in risky sexual behavior after ART initiation, by 2 or more years after ART, risky behavior had again risen. Although risky behavior eventually increased, there is no evidence that the increase among HIV-infected people was to levels higher than 2 or more years before ART initiation. There was some evidence to suggest that the availability of ART led to an increase of risky behavior among HIV-uninfected people, although this was inconsistent across different reported behavior variables. The HIV-uninfected population is substantially larger than the HIV-infected population. If risky behavior among this population increases due to the feeling that ART lowers the ‘death sentence’ that was once associated with HIV, then this change in behavior will affect the impact that ART has on the HIV epidemic. Policy makers are urged to intensify messages associating sexual behavior and HIV and to target both HIV-infected and uninfected people.
We thank the members of the Rural Clinical Cohort for their participation in the research, as well as the medical staff. This study was funded by the Medical Research Council of the UK.
Author contributions: L.A.S., R.N., K.O., and R.W. conducted or provided input on statistical analyses. L.A.S. wrote the manuscript. L.A.S., R.N., and R.C. conducted literature review. D.M. contributed to data interpretation and critically reviewed the paper. H.G. headed the MRC and L.V. and B.M. in turn directed the cohort from which data were provided. All authors contributed suggestions to the manuscript.
1. Cohen M. HIV transmission risk under ART. XVII International AIDS Conference; 2008.
2. Porco T, Martin JN, Page-Shafer KA, Cheng A, Charlebois E, Grant RM, Osmond DH. Decline in HIV infectivity following the introduction of highly active antiretroviral therapy. AIDS 2004; 18:81–88.
3. Scheer S, Chu PL, Klausner JD, Katz MH, Schwarcz SK. Effect of highly active antiretroviral therapy on diagnoses of sexually transmitted diseases in people with AIDS. Lancet 2001; 357:432–435.
4. Olley B. Higher-risk sexual behavior among HIV patients receiving antiretroviral treatment in Ibadan, Nigeria. Afr J AIDS Res 2008; 7:71–78.
5. Kennedy C, O'Reilly K, Medley A, Sweat M. The impact of HIV treatment on risk behavior in developing countries: a systematic review. AIDS Care 2007; 19:707–720.
6. Bunnell R, Ekwaru JP, Solberg P, Wamai N, Bikaako-Kajura W, Were W, et al. Changes in sexual behavior and risk of HIV transmission after antiretroviral therapy and prevention interventions in Uganda. AIDS 2006; 20:85–92.
7. Morgan D, Malamba SS, Maude GH, Okongo MJ, Wagner HU, Mulder DW, Whitworth JA. An HIV-1 natural history cohort and survival times in rural Uganda. AIDS 1997; 11:633–640.
8. Morgan D, Mahe C, Mayanja B, Okongo JM, Lubega R, Whitworth JA. HIV-1 infection in rural Africa: is there a difference in median time to AIDS and survival compared with that in industrialized countries? AIDS 2002; 16:597–603.
9. Mulder D, Nunn AJ, Kamali A, Nakiyingi J, Wagner HU, Kengeya-Kayondo JF. Two-year HIV-1-associated mortality in a Ugandan rural population. Lancet 1994; 343:1021–1023.
10. Seeley J, Wagner U, Mulemwa J, Kengeya-Kayondo J, Mulder D. The development of a community-based HIV/AIDS counselling service in a rural area in Uganda. AIDS Care 1991; 3:207–217.
11. Seeley J, Malamba SS, Nunn AJ, Mulder DW, Kengeya-Kayondo JF, Barton TG. Socioeconomic status, gender, risk of HIV-1 infection in a rural community in South West Uganda. Med Anthrop Quarterly 1994; 8:78–89.
12. Huber PJ. The behavior of maximum likelihood estimates under nonstandard conditions. Proc Fifth Berkeley Symp on Math Stats and Prob 1967; 1:221–233.
13. Liang K-Y, Zeger SL. Longitudinal data analysis using generalized linear models. Biometrika 1986; 73:13–22.
14. Zeger SL, Liang K-Y. Longitudinal data analysis for discrete and continuous outcomes. Biometrics 1986; 42:121–130.
15. Boerma T, Urassa M, Nnko S, Ng'weshemi J, Isingo R, Zaba B, Mwaluko G. Sociodemographic context of the AIDS epidemic in a rural area in Tanzania with a focus on people's mobility and marriage. Sex Transm Infect 2002; 78(Suppl 1):i97–i105.
16. Nnko S, Boerma JT, Urassa M, Mwaluko G, Zaba B. Secretive females or swaggering males? An assessment of the quality of sexual partnership reporting in rural Tanzania. Soc Sci Med 2004; 59:299–310.
17. Quigley M, Munguti K, Grosskurth H, Todd J, Mosha F, Senkoro K, et al. Sexual behavior patterns and other risk factors for HIV infection in rural Tanzania: a case-control study. AIDS 1997; 11:237–248.
18. Munguti K, Grosskurth H, Newell J, Senkoro K, Mosha F, Todd J, et al. Patterns of sexual behavior in a rural population in north-western Tanzania. Soc Sci Med 1997; 44:1553–1561.
19. Moatti J, Prudhomme J, Traore DC, Juillet-Amari A, Akribi HA, Msellati P, Côte d'Ivoire HIV Drug Access Initiative Socio-Behavioural Evaluation Group. Access to antiretroviral treatment and sexual behaviors of HIV-infected patients aware of their serostatus in Cote d'Ivoire. AIDS 2003; 17(Suppl 3):S69–S77.
20. Centers for Disease Control and Prevention (CDC). Increases in unsafe sex and rectal gonorrhea among men who have sex with men: San Francisco, CA, 1994–1997. Morb Mortal Wkly Rep 1999; 48:45–48.
© 2011 Lippincott Williams & Wilkins, Inc.